Geoscience Reference

In-Depth Information

Fig. 2.37
Creating correlatable shale bodies (shown in
blue
) in a fluvial system using SIS (Image courtesy of Simon

Smith)

When using the SIS method as commonly

applied in commercial packages, we need to be

aware of the following:

1. Reservoir data is generally statistically insuf-

ficient and rarely enough to derive meaningful

experimental variograms. This means that the

variogram used in the SIS modelling must be

derived by intuitive reasoning (see previous

section).

2. The range of the variogram is not the same as

the element body size. The range is related to

the
maximum
body size, and actual simulated

bodies can have sizes anywhere along the

slope of the variogram function. The range

should therefore
always
be set larger than

your expected average body size, as a rule of

thumb - twice the size.

3. The choice of the type of kriging used to start

the process off can have a big effect. For sim-

ple kriging a universal mean is used and the

algorithm assumes stationarity. For ordinary

kriging the mean is estimated locally through-

out the model, and consequently allows lateral

trends to be captured. Ordinary kriging works

well with large numbers of wells and well-

defined trends, but can produce unusual results

with small data sets.

4. Some packages allow the user to specify local

azimuths for the variogram. This information

can come from the underlying architectural

concept and can be a useful way of avoiding

the regular linear striping which is typical for

indicator models, especially those conditioned

to only a small number of wells.

2.7.2.3 Facies Trend Algorithms

The facies trend simulation algorithm is a

modified version of SIS which attempts to

honour a logical lateral arrangement of elements,